Skip to content

CryoSPARC on Kuma#

Step-by-step guide#

Installation:#

  1. Prior to the installation you will have to request a license_id on the obtaining a license ID page

  2. Send an email to 1234@epfl.ch with the title "HPC: CryoSPARC installation", and provide the following information: USERNAME (your GASPAR username or the username of a service account); ACCOUNT (your group's hpc account at SCITAS); CRYOSPARC_DEFAULT_ADMIN (a GASPAR username whose email will be used as the default admin. of the instance)

    Important

    1. Do not provide a License ID via the ticketing system.
    2. If the instance is to be used by multiple users, please ask the IT support of your faculty to create a service account with a lowercase username. Upper case username causes issue with the unix and slurm accounts. Using a service account helps protect personal data by ensuring that the shared instance is not associated with an individual's account and reducing the risk of exposing personal information through the instance..

Use:#

  1. Once the instance is installed by SCITAS, you will receive a message with instructions, including a link to the web interface for your CryoSPARC instance and the initial password, and another link to a shell for managing the instance with the access token.

  2. To manage your instance, use the shell generated for you. All commands (for example, cryosparcm status, stop, restart) work in the shell, except for upgrading the version, which should be done by SCITAS upon request. This is due to the way the CryoSPARC is installed as part as the docker image. If we would let you update the installation it will not persist after a restart of the container.

  3. Different lanes for different partitions are already created in the instance. The lanes are optimized for having maximum memory. Each lane works for both GPU and CPU-only jobs. For CPU-only jobs, lanes are configured to use the qos "build", which is limited to one job per instance (namely, you cannot submit more than one job at the same time and must wait for the submitted job completes before submitting another one). The default time for a job is set to 4 hours (4:00:00), and it can be changed in "Cluster submission script variables" for each lane.